IEEE TRANSACTIONS ON NEURAL SYSTEMS AND REHABILITATION ENGINEERING, VOL. 9, NO.

1, MARCH 2001

31

Quantification of the UPDRS Rigidity Scale

Susan K. Patrick, Allen A. Denington, Michel J. A. Gauthier, Associate Member, IEEE,
Deborah M. Gillard, Member, IEEE, and Arthur Prochazka, Senior Member, IEEE

AbstractIn the clinical setting, parkinsonian rigidity is

assessed using subjective rating scales such as that of the Unified
Parkinsons Disease Rating System (UPDRS). However, such
scales are susceptible to problems of sensitivity and reliability.
Here, we evaluate the reliability and validity of a device designed
to quantify parkinsonian rigidity at the elbow and the wrist. The
method essentially quantifies the clinical examination and employs small sensors to monitor forces and angular displacements
imposed by the clinician onto the limb segment distal to the joint
being evaluated. Force and displacement data are used to calculate
elastic and viscous stiffnesses and their vectorial sum, mechanical
impedance. Interexaminer agreement of measures of mechanical
impedance in subjects with Parkinsons disease was comparable
to that of clinical UPDRS scores. Examiners tended to overrate
rigidity on the UPDRS scale during reinforcement manoeuvres.
Mechanical impedance was nonlinearly related to UPDRS ratings
of rigidity at the elbow and wrist; characterization of such
relationships allows interpretation of impedance measurements in
terms of the clinical rating scales.

TABLE I
UPDRS RIGIDITY RATING SCALE (ADAPTED FROM [1])

Index TermsBiomedical measurements, clinical assessment,

force measurement, gyroscope, limb rigidity, Parkinsons diesease,
parameter estimation, reliability testing.

levels of interrater agreement, which in some cases were no

better than would be expected by chance [6][8]. Our own
observations have shown disagreement between clinicians as
to the effectiveness of medication [9], which may in fact be
explained by the tendency of different raters to concentrate to
different extents on the minimum, mean, or maximum rigidity
evoked. It has also been recognized that the severity of one
symptom may affect the assessment of another [unpublished
observations], [1], [10]. A need for more precise methods has
often been expressed [10][13].
Several groups have recognized the shortcomings of the subjective methods of assessment of rigidity and have developed
methods of quantifying, among other parameters, the work involved in moving a limb [14][16], or the stiffness [17], [18],
resonant frequency [19], or activation-induced increases in stiffness [20], [21] of the limb. Several of the earlier methods involved the use of torque motors which put unnatural constraints
on movement of the limb. Some of the more recent measurement
devices were designed to be used in ways that more closely resemble a clinical examination [21], [22]. The most often quoted
reasons for not introducing objective methods of quantification
into the clinical setting are the expense, complexity, and time
involved [1].
We recently introduced a device for the quantification of limb
stiffness which we applied to the measurement of parkinsonian
rigidity at the elbow [9]. This device is based on the method of
assessment of rigidity used routinely in the clinical setting, that
is, the passive manipulation of the joint in question by the examiner. Sensors are used to monitor the force imposed by the examiner in moving the limb as well as the amount of movement,
and the amount of rigidity is described in terms of mechanical
impedance as calculated from these two parameters. The device
is inexpensive, simple to use, and allows the examiner to feel and

I. INTRODUCTION
REATMENT of Parkinsons disease can be challenging
because of variation in symptoms between patients,
progression of the disease, and changing responses to medications [1][3]. The cardinal symptoms of Parkinsons disease
are rigidity, tremor, bradykinesia, and postural instability.
Rigidity responds well to levodopa and is one of the parameters
monitored to evaluate the efficacy of pharmacological and
surgical treatments. However, there is currently no standardized
objective method of measuring rigidity. At present, the clinician manipulates the limb of the patient and rates the evoked
stiffness according to an ordinal rating scale such as that of the
Unified Parkinsons Disease Rating System (UPDRS, Table I).
However, the subjective nature of such scales makes them open
to the interpretation of the examiner. Studies of the UPDRS
have found interrater reliability of the rigidity component to be
excellent [4], very good [2], and moderate [5]. Studies
of the Columbia University rating scale (on which the motor
component of the UPDRS is based), the Webster rating scale,
or a recently developed custom scale [6] have reported poorer

Manuscript received November 29, 1999; revised July 17, 2000 and December 8, 2000. This work was supported by the Alberta Heritage Foundation
for Medical Research, the Alberta Paraplegic Foundation, the Canadian MRC,
and a University of Alberta Dissertation Fellowship. A commercial party with
a direct financial interest in the results of the research supporting this article
has conferred, or will confer, a benefit upon one or more of the authors.
The authors are with the Division of Neuroscience, University of Alberta,
Edmonton, AB T6G 2S2, Canada (e-mail: spatric@emory.edu).
Publisher Item Identifier S 1534-4320(01)01782-X.

15344320/01$10.00 2001 IEEE

32

IEEE TRANSACTIONS ON NEURAL SYSTEMS AND REHABILITATION ENGINEERING, VOL. 9, NO. 1, MARCH 2001

thus qualitatively rate the evoked rigidity while it is also being

quantified. We found that a 50-second trial length provided a
more consistent estimate of mean stiffness than the customary
clinical exam, which typically might last about 10 s.
Any method of assessment of clinical symptoms must be
valid, sensitive to changes in the level of the symptom, and
reliable [2]. Since its introduction, the quantification device
has undergone substantial changes to the hardware. In this
paper, we test the sensors currently used, validate the algorithm
used to calculate stiffness, and assess reliability of the overall
approach. Since clinicians are accustomed to ordinal rating
scales, it may be preferable to express quantitative measures in
terms of such ratings. In this paper, we describe mathematically
the relationship between quantified stiffness measures and
UPDRS rigidity scores for the elbow and wrist and discuss
the appropriateness of using the resulting equations to express
quantified measures in terms of the UPDRS rating scale.
Parts of this paper have been presented elsewhere and appear
in abstract form [23][25].
II. METHODS
A. Stiffness Quantification Device
The device was originally designed to measure parkinsonian
rigidity at the elbow, and the original version has been described
in detail elsewhere [9]. A more recent version is reviewed here.
To quantify rigidity of a particular joint, the examiner repeatedly flexes and extends the joint as in a regular clinical examination. The movements are imposed through two air-filled pads
held distal to the joint (Fig. 1). The pads are connected to a differential force transducer (Motorola MPX10-DP) which measures the resultant force applied to the pads. A solid state piezoelectric gyroscope (Murata ENC05E) mounted on one of the
force pads monitors the imposed angular velocity from which is
computed displacement. (In the original version of the device, a
long elastic strain gauge was used to monitor displacement; the
gyroscope proved less cumbersome.) Both signals are fed into
a data acquisition box where they are filtered (9.6 Hz second
order low-pass Sallenkey), and digitized. A Motorola 68HC11
microprocessor samples the signals at 20 s and sends the result to a laptop computer. Alternatively, the filtered analog signals can be obtained directly from the box for external display
or sampling. In addition, a keypad on the box allows input of
clinical ratings or event markers.
Data collection was controlled via a user interface written
in Visual Basic (Microsoft). This program invokes MATLAB
(Version 4.2c.1, The MathWorks, Inc.) to perform the analysis.
(A more recent, commercially available version of the device
uses equivalent custom-written subroutines.) The first 0.5 s of
each trial, with the limb held still, allows the gyroscope signal
zero-velocity offset to be determined. The first four seconds of
all signals are discarded to avoid analyzing data collected before
the rigidity testing is actually underway. The gyroscope signal
is integrated using Eulers method and detrended. Calibration
of the raw data signals is then performed using linear interpolation. Calibration values are obtained by applying a 1 kg weight
to each force pad individually, and by rotating the gyroscope
through 90 degrees.

Fig. 1. Use of the rigidity quantification device. To quantify rigidity at the

elbow, the examiner flexes and extends the joint through two air-filled pads held
around the wrist. The pads are connected to a differential force transducer to
monitor the amount of force employed. A gyroscope mounted on one of the
pads is used to monitor angular velocity and displacement. A computer program
uses these data to calculate elastic stiffness, viscous stiffness, and mechanical
impedance. To quantify rigidity at the wrist, the forearm is pronated and the
pads held around the hand (not shown).

Calculation of stiffness values is performed by using a leastsquares parametric method to solve the following equation for
and over a 4-s moving window of data
(1)
where
torque measured;
angular displacement and angular velocity, respectively;
elastic stiffness;
viscosity;
constant offset of the sensors.
, where equals
the mean freThe viscous stiffness is
quency of the cyclical displacement. The output of the device
includes approximately 46-s traces of torque, angular displacement, and event markers, as well as approximately 42-s traces
and
. Mechanical impedance ( ) is the magnitude of
of
.
the vectorial sum of and
Unless otherwise specified, for all procedures described
below, raw force and gyroscope signals were taken directly
from the output of the data acquisition box (9.6 Hz second-order
low-pass SallenKey filter).
and

B. Gyroscope Validation
The performance of the gyroscope was tested against that of a
Penny and Giles goniometer (Biometrics, Ltd.) (P G), and that

PATRICK et al.: QUANTIFICATION OF THE UPDRS RIGIDITY SCALE

33

(CED 1401 and SIGAVG). A fast Fourier transform was

performed in MATLAB on the raw voltage signals; the frequency response and phase shift were determined by dividing
the cross spectral density by the power spectrum of the output
of the servo-motor force transducer.
D. Phase Shift Investigation

Fig. 2. Positioning of sensors on arm of subject for gyroscope validation. The

performance of the gyroscope in monitoring angular displacement of the arm
was compared to that of a Penny Giles goniometer (P G), and 3-D movement
analysis system (6D-Research). The gyroscope was placed along the ulna, just
distal to the elbow, and the P G goniometer was positioned along the elbow
on the ulnar side of the arm. Three electromagnetic sensors of the 6D system
were placed over the styloid process of the radius, and lateral epicondyle of the
humerus, and on the shoulder.

of an electromagnetic three-dimensional (3-D) movement analysis system (6D-Research, Skill Technologies, Inc.) (6D).
The sensors were affixed to the right arm of a human subject
with no known neurological impairment. The examiner (SP) applied flexion and extension of varying amplitude and frequency
to the elbow joint. The subject was asked either to relax or, occasionally, to provide some resistance. Positioning of each of
the sensors is shown in Fig. 2. Signals from the gyroscope and
P G goniometer were digitized at 20 s [Cambridge Electronic Design (CED) 1401 interface with 12 bit resolution, and
SIGAVG (version 5.42) software]. Data from the 6D electromagnetic sensors were sampled at 40 s using the 6D software.
Angular displacement was calculated off-line from the P G goniometer and integrated gyroscope signals using linear interpolation. The 6D system produced its own calculations of absolute joint angle based on vectorial projections. Calibrated signals were detrended and aligned to reduce any observed phase
shift, and the amplitudes of the signals were compared by calculating root mean square (RMS) errors.
C. Characterization of Force Sensors
As mentioned above, the force sensors of the device consist of
two air-filled pads connected to a differential force transducer.
To test linearity, combinations of weights ranging from 10 g to
4630 g were applied to one force pad at a time. Data were collected via the Visual Basic user interface and correlation coefficients between the forces applied and the force output signals
were computed.
To test the frequency response and phase shifting of the force
sensors, sinusoidal variations in force were applied to one of
the force pads using a custom-made moving-coil electromagnetic force servo-motor with a 4-strain-gauge proving-ring force
transducer. The frequency of the sinusoidal input (Feedback
Function Generator FG600) was increased from 0.2 to 10 Hz
over 60 s. The signal from the force sensor and the output of
the force transducer of the servo-motor were digitized at 100

Elastic stiffness ( ) is the ratio of the component of torque in

phase with angular displacement of the joint and the displacement itself; the viscous stiffness ( ) is the ratio of the component of torque in phase with angular velocity and the velocity
itself. The algorithm used to calculate mechanical impedance
to the
first estimates separately the contributions of and
torque measured. The mechanical impedance is then the vectorial sum of the two stiffnesses. Since any phase shifts of the
displacement or force signals with respect to each other will afand
calculated by the alfect the relative proportion of
gorithm, it is important that the sensors do not introduce phase
lags or leads.
This was tested for the force sensors as described above. To
test the gyroscopes, two gyroscopes were attached to the arm of
an angular servo motor (Printed Motors Ltd. Servalco servomotor) which moved through a range of 201 or 91.5 at frequencies ranging from 0.2 to 1.5 Hz. Signals from the gyroscope
and displacement transducers of the servo-motors were digitized
at 100 or 20 s (CED 1401 and SIGAVG). The displacement
transducer and integrated gyroscope signals were detrended and
calibrated by linear interpolation. The calibrated signals were
plotted and compared graphically in MATLAB.
E. Validation of System and TestRetest Reliability on a Model
Arm
In order to validate the results of the stiffness calculations,
we quantified the elastic stiffness of a model constructed from
a prosthetic arm and compared these measurements to the
estimates obtained using the rigidity quantification device.
The model is depicted in an inset of Fig. 3(a). Different levels
of stiffness were created by attaching different combinations
of elastic cords across the elbow. To obtain an independent
estimate of stiffness, the arm was flexed and extended via a
load cell at about 0.5 Hz over a constant range measured by
a goniometer. Angular displacement was measured using a
custom-built linear variable displacement transducer (LVDT).
This signal was low-pass filtered at 2.9 Hz (first order) and
10 Hz (second order SallenKey). LVDT and load cell signals
were digitized at 20 s (CED 1401 and SIGAVG). Five 50-s
tests were performed at each of four stiffness settings. The
stiffness was taken as the slope of the regression line of the plot
of angular displacement versus imposed torque.
After each set of five trials of a particular stiffness setting,
the stiffness was tested again at the same angular range and frequency for five trials using the rigidity quantification device.
The gyroscope was mounted on the ulnar side of the model
forearm, just distal to the elbow. For each trial, the mean elastic
stiffness was determined using two methods: 1) by the analysis software as described above; and 2) from the slope of the
regression line of the angle-torque plot. These two values were
then compared to the corresponding estimates of determined

34

IEEE TRANSACTIONS ON NEURAL SYSTEMS AND REHABILITATION ENGINEERING, VOL. 9, NO. 1, MARCH 2001

posthoc was used to test for significant differences between

the trials for each stiffness setting. When the data sets failed
tests of normality or equal variance, nonparametric versions
of the statistical tests were used (KruskalWallis ANOVA
and StudentNewmanKeuls test). Alpha was set at 0.01 for
ANOVA, and 0.05 for the posthoc tests.
F. Clinical Testing and Determination of Relationship of
Mechanical Impedance to Qualitative UPDRS Ratings

(a)

(b)
Fig. 3. Validation of calculations of elastic stiffness using a model arm. The
quantification device was tested on a prosthetic limb to which were attached
combinations of elastic cords to produce different levels of constant stiffness
[schematic inset of (a)]. Stiffness of the system was first estimated by repeatedly
flexing and extending the arm over a period of 50 seconds using a load cell and
monitoring the angular displacement with an LVDT. The elastic stiffness ( )
was taken as the slope of the regression line of the angle-torque plot of the two
signals [graph inset of (a)]. Stiffness of the system was subsequently measured
using the quantification device. Comparison of the estimates of calculated by
the software of the device to the values determined using the load cell and LVDT
yielded a regression coefficient of 0.999, and absolute values of were similar
(a). These data also offered a check of testretest reliability of the quantification
device on a system of constant stiffness. Shown are the mean of samples of Z
( SEM) of individual trials (b). Statistically significant differences between
some of the measurements of the same stiffness level are attributed to the small
standard errors, and are negligible on the UPDRS scale. The nonlinear equation
used to create the UPDRS axis from the axis was determined from the clinical
data in this paper (see Section III , part E).

K
K

from the load cell and LVDT using Pearsons correlation coefficient and linear regression.
These tests also allowed evaluation of the testretest reliability of the system, as five quantification trials were carried
data point
out at each level of stiffness. By taking every
trace (one point every four
of the mechanical impedance
seconds), 11 independent samples of were obtained for each
trial. One-way ANOVA with StudentNewmanKeuls test

Clinical testing of the device was performed in order to investigate the intra- and inter-rater reliability of measures of
produced by the device, and to compare this reliability to that
of the rigidity scale of the UPDRS. In addition, we wished to
compare the relationship between and the clinical ratings for
rigidity at the elbow to that determined using an older version
of the device [9], as well as to describe such a relationship for
rigidity at the wrist.
Four patients with idiopathic Parkinsons disease consented to
participateinthisstudyinaccordancewithlocalethicalcommittee
approval. The average age of the subjects was 58 years (range 52 to
64 years), and the average duration since onset of symptoms was
8 years (range 6 to 10 years). One subject (GM) had undergone a
right unilateral pallidotomy two years prior to this study. All subjects were tested without alteration of their daily medication routine; thus, at the time of testing, subjects were not necessarily receiving maximal pharmacological benefit.
Four examiners also participated in this study. JJ and WM are
neurologists, and MW is a physiotherapist; all three had extensive experience in clinical evaluation of parkinsonian rigidity.
AP, one of the authors of this paper, and one of the developers
of the quantification device, was less experienced in the clinical
evaluation of rigidity at the elbow. JJ and AP had both participated in a previous study involving the quantification device [9].
The rigidity of each patient was evaluated by each examiner.
The wrist and elbow of the right arm were assessed, separately,
in all cases. The examiner first rated the rigidity of the joints
using his or her regular method. Then one or two quantification trials lasting 50 s each were performed on each joint. The
examiners instructed the patient to perform a reinforcement manoeuvre of the examiners choice (e.g., clench contralateral fist,
or tap leg with contralateral hand) for roughly half of each trial.
For the quantification trials, the forearm was supinated while
the elbow was tested, and pronated while the wrist was tested.
Every few seconds over the course of each trial, as well as when
the rigidity level was felt to change, the examiner verbally rated
the rigidity according to the UPDRS; this rating was recorded
by SP using the keypad. For all subjects except one, the order
of examiners was: JJ, WM, MW, AP; for subject WD, the order
of JJ and WM was switched.
For purposes of statistical analysis, we assumed that independent samples of could be obtained from a single 50-s trial by
point). (Since
taking samples every four seconds (every
values imply that the subject was assisting the
negative or
movements imposed by the examiner, data points falling within
was negative were
a segment of the trace where either or
excluded; the point immediately after such a segment was selected instead.) The corresponding data points from the keypad
trace were used as independent samples of the clinical rating.

PATRICK et al.: QUANTIFICATION OF THE UPDRS RIGIDITY SCALE

The samples were grouped according to whether or not they

were collected during periods of reinforcement. Any samples
from nonreinforced segments occurring after a reinforced segment in the same trial were omitted, as effects of reinforcement
may still have been present.
and the clinical
To determine the relationship between
rating (UPDRS), the average of the samples of nonreinforced
or reinforced obtained from one trial was plotted against the
average of the corresponding samples of clinical rating. The
linear relationship was described as the slope of the regression
line passing through the origin and fitted to data from all
trials performed during the study. A nonlinear relationship was
also determined by fitting an exponential equation to the data
(SigmaPlot Version 4.01, Jandel Scientific).
Usually an examiner performed two trials of impedance quantification on each subject. Scores of nonreinforced or reinforced
rigidity ( or clinical rating) of one subject by one examiner
could be calculated by taking the average of all such samples
obtained from both trials. A combined rigidity score ( or clinical rating) was obtained by taking the average of the nonreinforced and reinforced scores; this allowed equal weighting of
either score so that the final measurement was not contingent
on the amount of time allotted to reinforcement maneuvers.
The Spearman rank-order correlation coefficient ( ) was
used as a measure of testretest reliability, by calculating
between the results of the first and second trials performed
on the same subject by the same examiner. Agreement between examiners was estimated by comparing the coefficient
of variation (standard deviation/mean) for each method of
measurement.
Impedance of five normal control subjects was measured by
SP using the quantification device. Subjects ranged in age from
46 to 76 years (mean 62.2 years) and included three males and
two females. Both wrists were tested in four subjects, and both
elbows in all five. Three 50-s tests were performed on each joint.
The subjects were asked to perform reinforcement manoeuvres
(tap the knee with the other hand) for roughly the last half of
the second and third trials. Impedance traces were subsampled
as described above. Scores of nonreinforced and reinforced
were calculated for each subject using the data from all trials
for one joint combined.
Unless otherwise stated, statistical analysis for all of the
above methods was performed using SigmaStat Version 1.0
(Jandel Scientific).
III. RESULTS
A. Gyroscope Validation
Passive angular displacement of the elbow joint of one subject
was measured using a gyroscope, P G goniometer and threedimensional movement analysis system (6D). The RMS error
of the gyroscope compared to P G goniometer (mean SEM
) or 6D system (
) represented less than
2.1
3% of the total range tested.
B. Characterization of Force Sensors
The linearity of the force sensors was tested by applying
weights of increasing magnitude to each of the force pads.

35

Correlation between the weight and the force output signal

was excellent, regression coefficients ( ) ranging from 0.997
to 1.000 for the trials using weight combinations ranging from
500 to 3630 g. Deviation from the regression line was 3% over
3630 g. Regression coefficients were equally good when
to
),
smaller weights were applied (10 to 500 g,
as well as when larger weights were investigated (1000 to
to
). The range of weights applied
4630 g,
corresponds to the forces used to passively move wrist and
elbow joints of control subjects or subjects with mild to severe
parkinsonian rigidity.
The frequency response and phase shift of the force sensors
was tested by applying sinusoidal variations in force to one of
the force pads via a linear servo-motor in force feedback mode.
Compared to the output of the proving- ring force transducer of
the servo-motor, the change in gain of the pad force sensor was
2 dB in the range of 0.2 to 10 Hz, and 0.5 dB over 0.2 to 2.0
Hz. The force pad signal lagged that of the force transducer of
the servo-motor strain gauge by 0.5 at 0.5 Hz, and 2 at 1 Hz.
This was opposite to the lead expected based on the frequency
response, and is thought to be caused by a delay due to the time
taken for the air to move through the tube connecting the force
pad to the differential force transducer. We conclude that the
force pad transducers are linear and have a flat frequency response over the range of frequencies of movements that would
be imposed during rigidity evaluation.
C. Phase Shift Investigation
Potential phase shift of the integrated gyroscope signal with
respect to the output of length transducers of servo-motors was
investigated.
When the gyroscope was tested over a 90 or 200 range
and its signal sampled at 20 s , the integrated signal (angular
displacement) generally lagged the output of the length transducer of the servo-motor by 20 to 30 ms (equivalent to about
4.5 phase lag at 0.5 Hz, and 9 phase lag at 1 Hz testing frequencies). The phase shift was negligible when the gyroscope
signal was sampled at 100 s , and this suggests that the observed phase shift at 20 s was a result of integration at a low
frequency of sampling and not an artifact of the sensor itself.
During evaluation of rigidity, the joint is generally tested over
an approximately 90 range, at a rate of 0.5 to 1 Hz. When
these ranges are considered, our results indicate that the relative
phase shift between the force and angular displacement signals
and
is small (4 to 7 ), and the effect on estimates of
minimal.
D. Validation of System and TestRetest Reliability on a Model
Arm
The stiffness calculations of the quantification device were
validated by comparing device-generated estimates of to the
stiffness of a model arm measured in separate tests with other
transducers [Fig. 3(a)]. The four stiffness settings tested covered the general range of elbow stiffnesses we have encountered
in patients with parkinsonian rigidity. There were excellent correlations between the independent estimates of obtained with
the LVDT and load cell and those computed by the device soft, Fig. 3(a)], or determined from angle-torque
ware [

36

IEEE TRANSACTIONS ON NEURAL SYSTEMS AND REHABILITATION ENGINEERING, VOL. 9, NO. 1, MARCH 2001

plots generated from the output of the device (

, not
shown). Absolute values were very similar; the slopes of these
regression lines were 0.983 and 0.954, respectively.
could be obtained
Since eleven independent samples of
from each trial that employed the quantification device, this
set of experiments presented an opportunity to evaluate the
testretest reliability of the device at each of four relatively
constant stiffness levels. Individual trials within a stiffness
setting were usually statistically different from each other
, and Stu(ANOVA or KruskalWallis ANOVA,
). However, the
dentNewmanKeuls test posthoc,
standard error of each test was very small; although the values
were different statistically, they were not different in a practical
sense given that the variability corresponded to a small fraction
of one point on the UPDRS scale [see Fig. 3(b)].
E. Clinical Testing and Relationship Between

and UPDRS

Four examiners measured the mechanical impedance of

the right elbow and wrist of four patients with idiopathic
Parkinsons disease using the quantification device. They also
rated the rigidity of the joints on the UPDRS scale both during
) as well as separately
the quantification trials (UPDRS
). Reinusing their usual method before the trials (UPDRS
forcement manoeuvres varied between examiners and subjects
and were chosen by the examiner. When the quantification
device was employed, the period during which the subject
performed reinforcement maneuvers ranged from one third of
the trial to the entire test period, as decided by the examiner.
One of the subjects (GM) experienced medication-induced
dyskinesias throughout the session.
ranged from 0.008 03 to 0.1058
Without reinforcement,
SEM: 0.0419
0.0219) for the elbow
N m/degree (mean
and 0.003 45 to 0.0254 N m/degree (mean SEM: 0.0114
0.004 98) for the wrist. Reinforcement produced overall sta(paired -test,
)
tistically significant increases in
for both joints ( with reinforcement: elbow: range 0.0178 to
SEM: 0.0638
0.0279, wrist:
0.1433 N m/degree, mean
range 0.008 90 to 0.0363 N m/degree, mean SEM: 0.0194
0.006 52).
When quantified measures ( ) of reinforced and nonreinforced rigidity from each trial performed at the elbow were
scores,
plotted against simultaneously-obtained UPDRS
the slope of the regression line fitting the data from all trials
was 27.4, giving a rough description of the relationship between
scores and the UPDRS rigidity scale. However, it is evident
from the plot that and clinical ratings are not linearly related
(Fig. 4). In attempt to describe this relation more accurately, the
data were fitted with an exponential equation in the form
UPDRS rating

(2)

To take into account a nonzero for control subjects, was set

to 0.0108 (mean of control subject elbow data, without reinforcement). Further, the parameter a was set to 4 to reflect the
maximum possible UPDRS rating. The result was the following
equation:
UPDRS rating

(3)

Fig. 4. Relationship between quantified measures and clinical ratings of

elbow rigidity. Quantified measures of mechanical impedance (Z ) were plotted
against simultaneously-assigned UPDRS ratings obtained during quantification
) showing a nonlinear relationship. The data were fitted by
trials (UPDRS
the line in the figure; the equation of this line was used as a means to express
Z in terms of the UPDRS rigidity scale. Each point represents the mean of
samples of nonreinforced (open symbols) or reinforced (filled symbols) rigidity
from one trial.

Using this equation, can be scaled so that and UPDRS

scores for individual subjects can be compared graphically.
and
Fig. 5(a) demonstrates the correlation between
scores shown in Fig. 4, and also indicates the
UPDRS
variation in scores for each subject. Each large point represents
the mean standard deviation of four scores (one mean score
from each examiner). The smaller circles represent the mean
scores of individual examiners for each subject. The combined
rigidity score of an examiner for a subject is the average of the
nonreinforced and reinforced rigidity scores; these are plotted
in Fig. 5(b). Combined scores are compared to the UPDRS
ratings, for which a single score takes both reinforced and
nonreinforced rigidity into account.
In general, interexaminer agreement for rigidity at the elbow
was greater with higher rigidity levels. Across subjects, the
mean ( SEM) coefficients of variation for raw values and
scores were 0.35 0.17 (range 0.160.84) and
UPDRS
0.59 0.22 (range 0.030.99), respectively, for nonreinforced
0.06 (range 0.200.44) and 0.33
0.20
rigidity and 0.31
(range 0.040.900), respectively, for reinforced rigidity. Taking
both nonreinforced and reinforced rigidity into consideration,
SEM coefficients of variation across subjects were
mean
0.06 (range 00.29) for UPDRS
scores, 0.32
0.17
0.16 (range 0.030.74) for combined UPDRS
scores,
and 0.28 0.08 (range 0.140.50) for raw values. Coefficients of variation were most consistent between subjects for
scores.
UPDRS
For combined rigidity scores, the maximum difference
for any subject was 1.5, comparable to the
in UPDRS
of 1.4. Combined
maximum difference between UPDRS
values scaled using (3) demonstrated a maximum intrasubject difference of 1.0. The average maximum difference in
scores across subjects was comparable for all three measures
).
(

PATRICK et al.: QUANTIFICATION OF THE UPDRS RIGIDITY SCALE

37

(a)

(b)
Fig. 5. Correlation between and interexaminer agreement of scaled Z and UPDRS scores of elbow rigidity. Raw Z measures were converted to the UPDRS scale
using the exponential in Equation (3) and compared to UPDRS ratings. Each large data point represents the mean standard deviation of four scores of rigidity
(one from each examiner); the size of the error bars reflects the level of interrater agreement. Small data points represent the mean score from individual examiners
for a subject. (a) Z (open circles) and UPDRS
(grey filled circles) measures for nonreinforced and reinforced rigidity are plotted separately. Note that there
scores (black filled circles) take into account both reinforced and
are no measures of nonreinforced rigidity for subject GM from examiner JJ. (b) UPDRS
nonreinforced rigidity, and are plotted with combined scores for scaled Z (open circles) and UPDRS
(grey filled circles) measures.

Testretest reliability at the elbow, as indicated by

Spearmans rank order correlation coefficients ( ) was
(raw )
greatest for combined rigidity scores [
)], when nonreinforced and reinor 0.94 (UPDRS
or
), or
forced rigidity scores were pooled (
when reinforced rigidity scores were considered separately
or
). Reliability was less for measures of
(
or
). Testretest relinonreinforced rigidity (
ability could not be estimated for the usual clinical method
) as each examiner rated the rigidity of a single
(UPDRS
joint only once.
score may offer a more accurate
It is proposed that the
description of changes of rigidity with reinforcement. In Fig. 4,
most of the points of reinforced rigidity (black points) lie
above the fitted line, suggesting a tendency of the examiners

to over-rate rigidity on the UPDRS scale when they know that

reinforcement manoeuvres are being performed.
In general, the observations from the measurements of
rigidity at the elbow hold for those at the wrist, although the
picture is not as clear with the wrist data. The plot of raw
values versus UPDRS
ratings yielded a linear regression
line with a slope of 76.2. As with the elbow, the relation
at the wrist is nonlinear and the
between and UPDRS
equation
UPDRS rating

(4)

was fitted to the data (Fig. 6).

Interexaminer agreement for clinical ratings (UPDRS
and UPDRS
) was generally better at higher levels of wrist

38

IEEE TRANSACTIONS ON NEURAL SYSTEMS AND REHABILITATION ENGINEERING, VOL. 9, NO. 1, MARCH 2001

the arm, resulting sometimes in resistance to the passive movements imposed. Reinforcement had no effect on impedance at
the wrist in control subjects. At the elbow, impedance was seen
to increase by as much as 273% with reinforcement, but this
was inconsistent; across subjects, reinforced impedance was
slightly higher but statistically insignificant than nonreinforced
impedance. Note that because negative values of
or
resulted in exclusion of data, the values presented above
represent the results from seven elbows (four subjects) and
seven wrists (four subjects). Data of reinforced impedance at
the wrist was only available from six joints (three subjects).

IV. DISCUSSION

Fig. 6. Relationship between quantified measures and clinical ratings of

wrist rigidity As with the elbow, an exponential equation (inset) was fitted
to the plot of raw Z values versus simultaneously-obtained UPDRS
scores of rigidity at the wrist. Each symbol represents the mean of samples of
nonreinforced or reinforced rigidity from one trial.

rigidity, whereas agreement between raw values appeared to

be independent of level of impedance. Coefficients of variation
scores ranged from 0.19 to
of raw values and UPDRS
0.65 (mean SEM 0.45 0.10) and from 0.37 to 0.80 (0.57
0.13), respectively, for nonreinforced rigidity, from 0.17 to
0.16) and from 0.08 to 0.77 (0.30
0.16), re0.73 (0.45
spectively, for reinforced rigidity, and from 0.13 to 0.69 (0.43
0.14) and from 0.11 to 0.77 (0.33 0.15), respectively, for
ratcombined scores. Coefficients of variation for UPDRS
ings ranged from 0.21 to 0.67 (0.39 0.10). Large variation
in during reinforcement for subject WR is attributed to two
things. First, examiner AP instructed a stronger reinforcement
manoeuvre than had the other examiners. Second, WR assisted
the limb movements imposed by examiner MW for most of the
two trials, the result being that only a few data samples were obtained from each trial. With the exception of subject WR, agreement of raw scores was comparable to or better than agreeratings.
ment between UPDRS
For combined wrist rigidity measures, the maximum difference in scores for a subject was comparable between
(1.5) and [scaled using (4), 1.7], and lowest for
UPDRS
(1.0). The average maximum difference across
UPDRS
(1.13), and similar
subjects was greatest for UPDRS
(0.76) and (scaled, 0.92)
between UPDRS
Spearman rank order correlation coefficients between
scores obtained from first and second trials at the wrist,
values
reflecting testretest reliability, were better for
to
) than for UPDRS
scores
(
to
). For quantified measures, coefficients
(
were equally good for nonreinforced as for reinforced and
ratings
combined rigidity scores; coefficients for UPDRS
were slightly lower with nonreinforced rigidity.
values for control subjects ranged from
Nonreinforced
0.0045 to 0.0216 N m/degree (mean SEM: 0.0108 0.0024)
for the elbow and 0.000 40 to 0.00264 N m/degree (0.00129
0.00028) for the wrist. Often subjects had difficulty in relaxing

The data presented here demonstrate that a device introduced

previously [9] provides a valid and effective means of quantifying rigidity of the upper limb.
The gyroscope, used to monitor angular displacement of the
limb, is accurate within 5%. The force sensors are linear with a
frequency response that is flat over the maximum range of frequencies of passive limb movement anticipated under the conditions of clinical evaluation of rigidity.
Results obtained from the quantification device were validated with those of another measure of stiffness on a model
arm. Trials performed using this model arm as well as clinical
trials on subjects with Parkinsons disease demonstrated good
testretest reliability of quantified measures.
In clinical trials, measures corresponded well to clinical
ratings of parkinsonian rigidity, both in individual trials and
across the data set as a whole. The slope of the regression line
data from the elbow (27.4) was close
fitting and UPDRS
to that obtained in our previous study (22.9) [9]. Two examiners
(AP, JJ) had participated in the earlier trials, but all subjects were
unique to the present study. In the current study, we have determined that the relationship between and clinical ratings is
best described with a nonlinear equation. This is in agreement
with the observations of Weber, Fechner, and Stevens that, in
general, perception is nonlinearly related to stimulus magnitude
[26]. While Fechner expressed the sensation-stimulus relationship in terms of a logarithmic function, Stevens later determined
that the relationship was better described by a power law, which
could be applied to noxious as well as neutral stimuli [26]. We
chose to use an exponential equation in order to accommodate
for the upper limit imposed on the UPDRS scale.
In our previous study, it was suggested that the slope of the regression line be used as a conversion factor with which to express
measurements in terms of UPDRS ratings. Though a linear conversion factor is simpler to understand and to apply, the exponential equation offers a more accurate picture of the relation between
and the UPDRS rigidity scale. There exists a nonlinear relationship between stiffness and its perception: what feels twice as stiff
actually is not. The exponential equation expresses the perception
of mechanical impedance better than a linear conversion.
Do (3) and (4) hold for all subjects? In our experience, the answer is a qualified yes, though in very slight subjects the clinical
rating may be somewhat underestimated by the equations (clinicians presumably take into account some estimate of a subjects

PATRICK et al.: QUANTIFICATION OF THE UPDRS RIGIDITY SCALE

potential strength). Taking into consideration that the analysis

procedure of the rigidity quantification device does not incorporate a correction for limb mass, and that linearity of measurement has been listed as an advantage of quantification systems
[21] (a joint twice as stiff produces a measure with twice the
magnitude), we suggest that rigidity measurements should eventually be reported in terms of Systme International (SI) units
rather than UPDRS ratings. In the mean time, the exponential
equations (3) and (4) provide a means whereby one may interpret, in general, the meaning of such measures with respect to
the clinically intuitive UPDRS scale.
It was observed that measurement of mechanical impedance
of the elbow of a completely relaxed control subject without hypertonus tends to produce negative values of . This is likely
due to the varying action of gravity on the mass of the forearm
as it is rotated. Our current method of discarding any segment
does not acof containing negative values of either or
commodate for situations of low stiffness with complete relaxation. Testing our model arm without elastic cords (i.e., no stiffness; the weight of this arm was comparable to that of a human
ranging from 0.0018 to
arm) produced negative values of
0.0083 N m/degree (mean
SEM: 0.0045
0.000 05).
This needs to be tested on human subjects showing no electromyographic activity of the muscles of the upper arm.
In the current study, interrater agreement and testretest reliabilityofquantifiedmeasuresofimpedanceweresimilartothoseof
UPDRS ratings of rigidity. The conclusion may then be drawn that
these measures are redundant and that clinical evaluation alone
would suffice. However, it is important to keep in mind that the
examiners involved in this study had worked together previously,
andthreeofthemhadbeeninvolvedinthefirstclinicalassessments
of the quantification device. Some studies have reported moderate
to excellent interrater reliability of the rigidity component of the
UPDRS scale, but in at least two of these studies comparisons
were made between examiners from the same center [4], [5], [27].
Another study evaluating the Columbia University and Webster
scales involved four examiners from different centers [8]. The
relatively poor reliability reported was attributed not only to the
inexperience of the raters, but also to differences in technique
employed and interpretation of the scales. Another study evaluating these same two scales with six neurologists from the same
center reported only fair to moderate interrater reliability [7],
perhaps an indication of shortcomings in the scales themselves. In
any case, differences in the method of rigidity assessment is one
source of disagreement between clinicians.
In an unpublished clinical trial of our quantification device
at the National Hospital, Queen Square, London, U.K., five
neurologists from three countries assessed rigidity at the elbow
and wrist of seven subjects with Parkinsons disease. Their
UPDRS ratings for a given patient differed by as much as
two points. However, their ratings subsequently collected over
) tended to
50-second trials with the device (UPDRS
show better agreement. The improvement was apparently due
to specification of the plane of movement of the joint and
prolongation of the length of the test. The need for uniformity
in methods of application of rating scales has long been recognized [1], [28], [29]. It has been recommended that clinicians
involved in a study establish agreement on the technique to be

39

used and interpretation of scales [8], [30]. However, this does

not remedy the problem of comparing results between studies.
Quantification of rigidity can help by removing the subjective element of the scales. In the London study above, persisting
interrater discrepancies between UPDRS
scores obtained
under the more controlled conditions of quantification of
rigidity were attributed largely to individual interpretation of
the scale. In the present study, it is thought that the quantified
measures more accurately reflected increases in rigidity with
reinforcement maneuvers.
It is important to note that standardization of technique is also
desirable when a quantification device is employed, since many
variables can affect the level of muscle tone observed [1]. In the
present study, occasional large variability in measures of mechanical impedance was most likely due to actual differences
in the level of rigidity evoked. For example, examiner AP instructed subject WR to perform a reinforcement manoeuvre that
was stronger than requested by the other examiners.
In this paper, we have demonstrated validity and reliability of
a device for the quantification of parkinsonian rigidity. The device fulfills the requirements demanded by a clinical setting, that
is, it is relatively inexpensive, time-efficient, and simple. The
way in which this device is used more closely emulates the clinical examination than other methods of quantitative assessment.
Measurements are sensitive to changes in impedance levels due
to reinforcement, as well as to fluctuations in rigidity which
occur spontaneously during the course of a trial and which are
detectable by the clinician. The nonlinear relationships between
and simultaneously-acquired UPDRS scores of rigidity at the
elbow and wrist were described using exponential equations.
Because differences in technique of rigidity assessment and interpretation of rating scales appear most evident when clinicians
from different centers are compared, a study involving a large,
international sample of clinicians would be desirable to validate
and possibly refine these equations.
The quantification device evaluated here is currently in use
in several clinical and research settings in North America and
Europe and the number is slowly growing. If the device achieves
widespread acceptance, then the choice will need to be made as
to whether to communicate the measures of rigidity in terms of
mechanical impedance or as scores relating to the UPDRS scale.
Scientifically, the former approach is preferable, but it may be
that the method will gain better initial acceptance if measures
are interpreted in terms of the familiar clinical rating scales.
in terms of UPDRS ratings using equations
Expression of
such as those determined here may subsequently allow a gradual
transition to the expression of rigidity in the standard physical
units of the Systme International (SI).

ACKNOWLEDGMENT
The authors would like to thank J. H. Jhamandas, W. R. Martin,
and M. Wieler for their valuable participation in this study. They
would also like to express appreciation for the involvement of J.
C. Rothwell, P. Limousin, F. Yokoshi, C. D. Marsden, N. Quinn,
and P. Brown in an earlier clinical trial of the quantification device
at the Institute of Neurology, National Hospital, Queen Square,

40

IEEE TRANSACTIONS ON NEURAL SYSTEMS AND REHABILITATION ENGINEERING, VOL. 9, NO. 1, MARCH 2001

London, U.K. In addition, they would like to thank D. J. Bennett

and K. Yoshidafor theirexpertiseand help with data analysis.

REFERENCES
[1] A. E. T. Lang and S. Fahn, Assessment of Parkinsons disease, in
Quantification of Neurological Deficit, T. L. Munsat, Ed. Boston, MA:
Butterworths, 1989, pp. 285309.
[2] P. Martnez-Martn, Rating scales in Parkinsons disease, in
Parkinsons Disease and Movement Disorders, 2nd ed, J. Jankovic and
E. Tolosa, Eds. Baltimore, MD: Williams and Wilkins, 1993, pp.
281292.
[3] C. D. Marsden and M. Schachter, Assessment of extrapyramidal disorders, Br. J. Clin. Pharmacol., vol. 11, no. 2, pp. 129151, 1981.
[4] J. M. Rabey, H. Bass, U. Bonuccelli, D. Brooks, P. Klotz, A. D. Korczyn, P. Kraus, P. Martnez-Martn, P. Morrish, W. Van Sauten, and B.
Van Hilten, Evaluation of the short Parkinsons evaluation scale: A new
friendly scale for the evaluation of Parkinsons disease in clinical drug
trials, Clin. Neuropharmacol., vol. 20, no. 4, pp. 322337, 1997.
[5] M. Richards, K. Marder, L. Cote, and R. Mayeux, Interrater reliability
of the unified Parkinsons disease rating scale motor examination, Mov.
Disord., vol. 9, no. 1, pp. 8991, 1994.
[6] L. R. Van Dillen and K. E. Roach, Interrater reliability of a clinical scale
of rigidity, Phys. Ther., vol. 68, no. 11, pp. 16701681, 1988.
[7] A. Ginanneschi, F. DeglInnocenti, S. Magnolfi, M. T. Maurello,
I. Catarzi, P. Marini, and L. Amaducci, Evaluation of Parkinsons
disease: Reliability of three rating scales, Neuroepidemiology, vol. 7,
no. 1, pp. 3841, 1988.
[8] G. Geminiani, B. M. Cesana, F. Tamma, P. Contri, C. Pacchetti, F.
Carella, R. Piolti, E. Martignoni, P. Giovannini, F. Girotti, and T.
Caraceni, Interobserver reliability between neurologists in training of
Parkinsons disease rating scales: A multicenter study, Mov. Disord.,
vol. 6, no. 4, pp. 330355, 1991.
[9] A. Prochazka, D. J. Bennett, M. J. Stephens, S. K. Patrick, R. SearsDuru, T. Roberts, and J. H. Jhamandas, Measurement of rigidity in
Parkinsons disease, Mov. Disord., vol. 12, no. 1, pp. 2432, 1997.
[10] C. D. Ward, J. N. Sanes, J. M. Dambrosia, and D. B. Calne, Methods
for evaluating treatment in Parkinsons disease, Adv. Neurol., vol. 37,
pp. 17, 1983.
[11] H. Tervinen, J. K. Tsui, and D. B. Calne, Evaluation of Parkinsons
disease, in Drugs for the Treatment of Parkinsons Disease, D.
B. Calne, Ed. Berlin, Germany: Springer-Verlag, 1989, vol. 88,
Handbook of Experimental Pharmacology, pp. 271279.
[12] W. G. Ondo, J. Jankovic, E. C. Lai, C. Sankhla, M. Khan, L. Ben-Arie,
K. Schwartz, R. G. Grossman, and J. K. Krauss, Assessment of motor
function after stereotactic pallidotomy, Neurology, vol. 50, no. 1, pp.
266270, 1998.
[13] J. A. Obeso, G. Linazasoro, J. C. Rothwell, M. Jahanshahi, and R.
Brown, Assessing the effects of pallidotomy in Parkinsons disease
[letter], Lancet, vol. 347, no. 9013, p. 1490, 1996.
[14] D. D. Webster, A method of measuring the dynamic characteristics of
muscle rigidity, strength, and tremor in the upper extremity, IRE Trans.
Med. Electron., pp. 159164, Sept. 1959.
[15] H. Tervinen, J. K. C. Tsui, E. Mak, and D. B. Calne, Optimal indices
for testing parkinsonian rigidity, Canad. J. Neuro. Sci., vol. 16, no. 2,
pp. 180183, 1989.
[16] C. Kirollos, A. Charlett, C. J. A. ONeill, R. Kosik, K. Mozol, A.
G. Purkiss, S. G. Bowes, P. W. Nicholson, W. B. Hunt, C. Weller, S.
M. Dobbs, and R. J. Dobbs, Objective measurement of activation
of rigidity: Diagnostic, pathogenic and therapeutic implications on
parkinsonism, Br. J. Clin. Pharmacol., vol. 41, no. 6, pp. 557564,
1996.
[17] J. Ghika, A. W. Wiegner, J. J. Fang, L. Davies, R. R. Young, and J.
H. Growdon, Portable systems for quantifying motor abnormalities in
Parkinsons disease, IEEE Trans. Biomed. Eng., vol. 40, pp. 276283,
Mar. 1993.
[18] A. W. Wiegner and R. L. Watts, Elastic properties of muscles measured
at the elbow in manI: Normal controls, J. Neurol. Neurosurg. Psych.,
vol. 49, no. 10, pp. 11711176, 1986.
[19] M. Lakie, E. G. Walsh, and G. W. Wright, Resonance at the wrist
demonstrated by the use of a torque motor: An instrumental analysis
of muscle tone in man, J. Physiol., vol. 353, pp. 265285, Aug. 1984.
[20] M. P. Caligiuri and D. R. Galasko, Quantifying drug-induced changes
in parkinsonian rigidity using an instrumented measure of activated stiffness, Clin. Neuropharmacol., vol. 15, no. 1, pp. 112, 1992.

[21] M. P. Caligiuri, Portable device for quantifying Parkinsonian wrist

rigidity, Mov. Disord., vol. 9, no. 1, pp. 5763, 1994.
[22] D. Halpern, R. Patterson, R. Mackie, W. Runck, and L. Eyler, Muscular
hypertonia: Quantitative analysis, Arch. Phys. Med. Rehabil., vol. 60,
no. 5, pp. 208218, 1979.
[23] S. K. Patrick, S. C. Naaman, M. J. A. Gauthier, and A. Prochazka, A
device for the quantification of Parkinsonian rigidity [abstract], in Society for Neuroscience, 1995.
[24] S. K. Patrick, A. Denington, M. J. A. Gauthier, and A. Prochazka,
Sensing limb displacement to quantify limb stiffness [abstract], in
Society for Neuroscience, 1997.
[25] S. K. Patrick and A. Prochazka, Testing of a device for the quantification of limb stiffness [abstract], Can. J. Physiol. Pharmacol., vol. 74,
p. Axxv, 1996.
[26] J. G. Snodgrass, G. Levy-Berger, and M. Haydon, Human Experimental
Psychology. New York: Oxford Univ. Press, 1985.
[27] P. Martnez-Martn, A. Gil-Nagel, L. M. Gracia, B. Gmez, J. MartnezSarris, F. Bermejo, and T. C. M. Group, Unified Parkinsons disease
rating scale characteristics and structure, Mov. Disord., vol. 9, no. 1, pp.
7683, 1994.
[28] C. C. Goetz, G. T. Stebbins, T. A. Chmura, S. Fahn, H. L. Klawans,
and C. D. Marsden, Teaching tape for the motor section of the unified Parkinsons disease rating scale, Mov. Disord., vol. 10, no. 3, pp.
263266, 1995.
[29] J. W. Langston, H. Widner, C. G. Goetz, D. Brooks, S. Fahn, T. Freeman,
and R. Watts, Core assessment program for intracerebral transplantations (CAPIT), Mov. Disord., vol. 7, no. 1, pp. 213, 1992.
[30] G. K. Montgomery, N. C. Reynolds, and R. M. Warren, Qualitative assessment of Parkinsons disease: Study of reliability and data reduction
with an abbreviated Columbia scale, Clin. Neuropharmacol., vol. 8, no.
1, pp. 8392, 1985.

Susan K. Patrick received the B.Sc. degree in general science, and the Ph.D.
degree in neuroscience from the University of Alberta, Edmonton, AB, Canada,
in 1992 and 1999, respectively.
Her research involves development of methods to quantify tone in the upper
limb in Parkinsons disease and hemiplegia, and application of these quantification methods to the clinical and surgical setting.

Allen A. Denington received the Dip. in electronics engineering technology

from the Northern Alberta Institute of Technology, Edmonton, AB, Canada, in
1994.
Since December 1994, he has been employed by the Division of Neuroscience
at the University of Alberta. He is involved in the design, development, miniaturization, and production of medical electronic devices, laboratory electronics
and mechanisms. His work includes the production of battery powered microprocessor controlled muscle stimulators.

Michel J. A. Gauthier (S87A89) received the B.Sc. degree in mechanical

engineering from the University of Alberta, Edmonton, AB, Canada, and the
Dip. computer engineering technology from Westerra Institute of Technology,
Stony Plain, AB, Canada, in 1985, and 1989, respectively.
Since 1989, he has been with the Division of Neuroscience at the University
of Alberta. His work includes the development of microprocessor controlled
muscle stimulators, laboratory electronics and mechanisms, and software and
firmware for analysis and control.

Deborah M. Gillard (S91M92) received the B.Sc. and M.Sc. degrees in electrical engineering from the University of Alberta, Edmonton, AB, Canada, in
1991 and 1994, respectively.
Since 1995, she has been with the Division of Neuroscience at the University of Alberta as a Research Associate. She has been involved in the design,
development, and testing of electrical stimulation devices to improve functional
movement. Her areas of interest include system identification and adaptive control.

PATRICK et al.: QUANTIFICATION OF THE UPDRS RIGIDITY SCALE

Arthur Prochazka (M74SM78) received the B. Eng. and the M.Sc. degrees
from the University of Melbourne, Australia, and the Ph.D. degree from the
University of Ulm, Germany, in 1967, 1970, and 1973, respectively.
After postdoctoral studies at Monash University, Melbourne, he moved to the
U.K. in 1977 as a Lecturer of Physiology, St. Thomass Hospital, University of
London. In 1986, he was appointed Professor of Physiology at the University of
Alberta, Edmonton, AB, Canada. His research has focused on the sensorimotor
control of mammalian movement and, in recent years, the development of electronic devices to measure and improve movement in neurological disorders.

41